Mining physiological signals for motion sickness identification

The incongruity between the movement recognized by the vestibular system and the one recognized by the visual system can generate motion sickness. Taking a journey by car is one of the contexts in which this phenomenon is manifested. It is therefore interesting for automotive companies to try to automatically reveal this phenomenon to remedy it and improve the experience of people in the passenger compartment. This thesis work presents a data analysis solution to automatically analyse the physiological response of a subject suffering from motion sickness. The goal is to early detect the onset of motion sickness in order to implement timely corrective actions that can stop the discomfort in the bud. In this thesis work, the physiological response of the subject is analyzed considering the level of sweating of the subjects examined. This level of sweating is recorded as dermal conductivity. In the proposed solution, the target signal has been analyzed to extract a set of features modeling the signal trend. The extraction of features from the physiological signal takes place through sliding windows, to have information on the temporal evolution of the signal. Each window is therefore associated with a set of features and a class label indicating whether the subject suffered from motion sickness or not in the time interval associated with the window. The class label is defined based on the information provided by the subject itself during the experiment. Different classification algorithms have been explored and compared to build a classification model able to automatically identify the motion sickness onset. The experimental evaluation showed that the proposed solution provides interesting and promising results.